Documentation
Everything you need to route well.
OpenSmartRoute routes each request to a model, agent, skill, tool or person and returns the decision with the trace that produced it. Use it as a hosted platform, as a Python package, as a self-hosted service, or through the browser extension and the desktop app. Pick the way in below; every page names the endpoint, command or setting it describes.
v2.1.0Released 2026-10-15.Release notes
Apps
The router where you already write: a card under the composer of the chat sites, and one shortcut over every application on your computer.
Hosted platform
Send requests to the routing API with an API key, or point an OpenAI client at /v1 with model "auto".
Python SDK
Install the package, describe targets in YAML and route in-process. No runtime dependencies.
Self-hosting
Run osr serve as a container or Helm release with the same /route, /feedback and /v1 endpoints, hardened with the guard, health breakers, audit and a shared state store.
Install the command line and sign in
One command; the Python SDK comes with it.
curl -LsSf https://opensmartroute.ai/install.sh | sh
osr loginpowershell -ExecutionPolicy ByPass -c "irm https://opensmartroute.ai/install.ps1 | iex"
osr loginAlready have a Python toolchain? uv tool install "opensmartroute[yaml,server]" or pip install "opensmartroute[yaml,server]" work too. osr login opens the browser and stores a workspace API key for this platform; self-hosted routers accept osr login --url http://router:8000 --token .... From TypeScript or JavaScript: npm install @opensmartroute/sdk (TypeScript SDK).
Get started
What OpenSmartRoute is and the fastest path to a first routed request - hosted or in your own code.
5 pages
- What is OpenSmartRouteThe open-source AI model router: what it routes, how it decides, what it costs to run and where to start.
- Platform quickstartFive minutes to a first routed request: workspace, API key, POST /api/v1/route, the OpenAI-compatible endpoint, feedback.
- Python SDK quickstartFive minutes to routing in your own process: install, describe targets, route, constrain, learn from outcomes.
- Install the platformRun the whole platform yourself: Azure Marketplace or Docker Compose, first sign-in, custom domain, upgrade and backup.
- Use cases by industryNine complete, working catalogues - healthcare, banking, legal, retail and more - with the requests users send and where each one routes.
Hosted platform
The routing API behind an API key: decisions with trace, the OpenAI-compatible endpoint, plans and quotas.
8 pages
- Platform guideAuthentication, POST /api/v1/route, the OpenAI-compatible endpoint, feedback, plans, tenants, dashboard and errors.
- Platform journeysWhat each plan unlocks and where it lives; the first day step by step, everyday tasks by role, plan changes and offboarding.
- Model serverHost open-weight models on your own hardware behind the Ollama and OpenAI APIs: pulls, residency with a fit check, admission queue, metrics; CPU and CUDA.
- TypeScript SDKRouting, feedback, estimates, streaming chat and every workspace call from Node, Deno, Bun, edge runtimes and browsers.
- MarketplaceFind, install, buy and publish agents, skills, personas, prompts and stack templates; ratings and the review lifecycle.
- Cost estimates and the MCP serverQuote a request before sending it, recommended models per use case, and the MCP server that puts the router in VS Code, Cursor, Claude and any agent.
- Content publishing for your websiteYour feeds and briefs become posts written through the router, reviewed in the dashboard and delivered to your site - a signed webhook, WordPress or Ghost - or pulled as RSS; the same tools on your MCP server.
- REST API referenceEach endpoint with parameters, request and response schemas and an example request.
Apps
The router where people write: the browser extension under the composer of the chat sites and in Google Docs, the desktop bar over every application, the editor panel in VS Code, the pane in Word.
4 pages
- Browser extensionThe pre-send advisor for ChatGPT, Claude, Gemini, Copilot and Perplexity: risks, redaction, model hints and cost quotes on the device.
- Desktop appOne shortcut over any app on Windows, macOS and Linux: the routed command bar reads your selection and pastes the answer back.
- VS Code extensionSelect code or text in the editor: task, language, size, personal data, the recommended model and its cost - then ask, explain or rewrite in place through the router.
- Word add-inSelect a passage in Word: task, language, size, personal data, the recommended model and its cost - then ask, explain, summarise, rewrite or proofread through the router.
Python SDK
The router as a library: targets, rules, decorators, the enterprise builder and the public API.
9 pages
- User guideThe complete walkthrough: targets, rules, plans, real providers, execution, learning, the CLI and SKILL.md packages.
- SDK guideDecorators for strategies, rules, signals, middleware and telemetry; settings, error handling, threading and plugins.
- Enterprise builderRouterBuilder: auto-learning, health, budgets, tenants, audit, stores and rollout.
- Compatibility matrixEvery provider, client, agent framework and adapter the router connects to - and how each connection was verified.
- Tracing and observabilitySpans and events for every routing stage; memory, metrics, log, file and OpenTelemetry sinks; /events, /trace and /metrics.
- Agent engineeringPrompt, harness, context, memory and agent engineering, integrations, guardrails and evals - where each lives, and the Langfuse, Phoenix, Ragas and DeepEval connections.
- Benchmarks and leaderboard recipeCollect the public preference suites, train the routing SLM and publish a reproducible benchmark run.
- Published benchmark runsThe published run on MT-Bench human, PPE human and WebDev Arena against the declarative router and every baseline.
- Python API referenceEvery public module, class and function with its signature, generated from the source.
Self-hosting
Run the same router as a service on your own infrastructure.
1 page
Concepts
How a request becomes a decision, the mathematics behind each score and the research the modules draw on.
4 pages
- ArchitectureHow a request becomes a decision: signals, policy, strategies, utility, plans.
- MathematicsThe utility function, calibration, IRT, Bradley-Terry, LinUCB and Markov derivations.
- ResearchProblem statement, literature by family and which module borrows which idea.
- Open Capability ManifestThe OCM specification and JSON Schema for describing routable capabilities.
Security
Threat model, hardening and how to report a vulnerability.
3 pages
- Security modelThreat model, guard middleware, PII handling, provenance and the red-team suite.
- External review packScope, trust boundaries, the evidence the project produces and the review log for an independent reviewer.
- Reporting vulnerabilitiesSupported versions and how to report a security issue privately.
Agent skills
Agent-Skills packages (SKILL.md) for coding assistants working with OpenSmartRoute; each one is also a valid routing target.
7 pages
- osr-decorator-sdkExtend OpenSmartRoute in Python with the decorator SDK.
- osr-deploy-serveRun OpenSmartRoute as a service.
- osr-enterprise-builderAssemble a production OpenSmartRoute router with RouterBuilder.
- osr-evaluationEvaluate and benchmark an OpenSmartRoute router.
- osr-integrationsConnect OpenSmartRoute to the outside world.
- osr-routing-catalogueDefine and tune an OpenSmartRoute routing catalogue.
- osr-security-hardeningHarden OpenSmartRoute deployments against prompt injection, gadget/steering attacks, PII leakage, runaway agents and untrusted tool arguments.
Releases
What changed in each version, and how versions and release channels work.
2 pages